Skip to content

ci,op-acceptance-tests: run SafeDB op-geth resync tests in dedicated jobs - #23824

Closed
einar-oplabs wants to merge 3 commits into
einar/safedb-remaining-validationfrom
einar/safedb-resync-job
Closed

einar-oplabs wants to merge 3 commits into
einar/safedb-remaining-validationfrom
einar/safedb-resync-job

Conversation

@einar-oplabs

Copy link
Copy Markdown
Contributor

Follow-up to #23787, splitting the op-geth SafeDB resync step out of the sharded acceptance jobs as agreed in #23787 (comment).

The two resync cases SkipOnOpReth, so the sharded op-reth suite never exercises them. #23787 ran them as an extra step on shard 0 of each acceptance job, which shard-tests.sh cannot account for, so shard 0 was always longer by the full resync runtime. This moves the step into an op-acceptance-tests justfile recipe, safedb-resync-test, and runs it in two dedicated single-node jobs, safedb-resync-opn-op-geth-fusaka and safedb-resync-kona-op-geth-fusaka, on xlarge.gen2. Both are required by ci-gate, as the shard-0 step was before. The recipe keeps the -run restriction and the fail-on-skip guard, forces DEVSTACK_L2EL_KIND=op-geth so it is also runnable locally, and writes the same logs/<run>/summary.log layout as acceptance-test so the job's result-printing steps work unchanged.

Based on einar/safedb-remaining-validation; retarget to develop once #23787 merges.

Validation:

  • Merged CircleCI config validates with circleci config validate --org gh/ethereum-optimism; all 23 ci-gate requires resolve to workflow job names.
  • just --list and just -n safedb-resync-test render; ShellCheck on the rendered body reports only the justfile's own backtick REPO_ROOT constant, shared with acceptance-test.
  • go test -list with the recipe's run filter selects exactly TestPreserveDatabaseOnCLResync and TestTruncateDatabaseOnELResync.
  • CI: both new jobs green; shard 0 of the memory-all jobs no longer runs the resync step.

🤖 Generated with Claude Code

einar-oplabs and others added 3 commits October 30, 2026 19:36
…consumers

Validate SafeDB history across restart and resync, Cannon execution-step
proofs through game resolution, and dispute-monitor proof consumers.
Run op-geth resync coverage sequentially before normal acceptance tests
for both consensus clients, and document the historical RPC contract.
…dy consumer tests

The op-geth SafeDB resync step now runs after the sharded suite so a resync
failure cannot skip shard 0's share of it, restricts itself to the two cases
that SkipOnOpReth, and fails when any of them skips or does not pass.

The cannon_kona preset relies on the runtime's default SafeDB path, the
dispute-monitor classification test drops a wait on a status new games already
have, and the CL-sync resync comment describes CL sync rather than EL sync.

The safeHeadAtL1Block and superroot_atTimestamp docs gain the invocation table
and example style used by the other methods on the page.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
…jobs

The two resync cases SkipOnOpReth and ran as an extra step on shard 0 of
each sharded acceptance job, which shard-tests.sh cannot balance. Move the
step into a safedb-resync-test justfile recipe and run it in its own
single-node op-node and kona-node jobs, both required by ci-gate.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
@einar-oplabs einar-oplabs self-assigned this Nov 1, 2026
@einar-oplabs
einar-oplabs requested a review from joshklop November 1, 2026 20:25
@einar-oplabs
einar-oplabs marked this pull request as ready for review November 1, 2026 20:26
@einar-oplabs
einar-oplabs requested a review from a team as a code owner November 1, 2026 20:26
@einar-oplabs
einar-oplabs requested a review from aakoshh November 1, 2026 21:01
@einar-oplabs
einar-oplabs force-pushed the einar/safedb-remaining-validation branch from 200d261 to 269d763 Compare November 11, 2026 13:32
@einar-oplabs
einar-oplabs requested a review from a team as a code owner November 11, 2026 13:32
@sebastianst

Copy link
Copy Markdown
Member

🤖 Fable says: A note from the op-geth decoupling work (#20257). This PR adds two ci-gate-required jobs (safedb-resync-{opn,kona}-op-geth-fusaka) and a justfile recipe that force DEVSTACK_L2EL_KIND=op-geth.

op-geth is EOL as of Karst, and #21744 removes the in-process op-geth L2 EL from op-devstack/sysgo (l2_el_opgeth.go, MixedL2ELOpGeth, DEVSTACK_L2EL_KIND). It cannot exist after the go.mod flip to upstream go-ethereum (#20266): the same code would silently run upstream geth as the L2 EL. So please don't add new test or CI dependencies on op-geth; they would have to be removed again before the flip.

#23827 makes the SafeDB resync tests run on op-reth in the existing acceptance jobs, which makes these jobs unnecessary. Can this be closed in favour of #23827?

@einar-oplabs

Copy link
Copy Markdown
Contributor Author

Opus 5.5 (high): Superseded. #23827 runs the two SafeDB resync tests in the normal op-reth acceptance jobs, and #23787 dropped its op-geth step in 68cf64e, so there is no op-geth step left to move into dedicated jobs. op-geth is also being removed from op-devstack (#20257, #21744).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants